From an LDV perspective, SFDC mentions Divisions are a way to speed up query response times:
Divisions... make searches, reports, and list views more meaningful to users in very large organizations by segmenting data into logical sections. For example, if your searches are taking a long time or timing out, and your users only need to see a subset of your data, then using Divisions can help improve performance.
I understand the user-experience when using Divisions... simple enough. Useful. But technically, how does this feature work?
Is this a prioritized custom index that the platform maintains?